Skip to content
View hdprajwal's full-sized avatar
16 followers · 12 following

Achievements

Achievement: QuickdrawAchievement: Pull Sharkx2Achievement: Pair ExtraordinaireAchievement: YOLOAchievement: Arctic Code Vault Contributor

Achievements

Achievement: QuickdrawAchievement: Pull Sharkx2Achievement: Pair ExtraordinaireAchievement: YOLOAchievement: Arctic Code Vault Contributor

Highlights

  • Pro

Block or report hdprajwal

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
hdprajwal/README.md

Hi, I'm Prajwal

AI engineer building agent tooling and applied ML, with three years of backend and full-stack at Opslyft.

Publication

Projects

  • BaitBench: a benchmark that checks how often an LLM, writing a job application for you, falls for a prompt injection hidden in the job post. It scores models on a resistance leaderboard, live at baitbench.hdprajwal.dev.
  • emberd: Firecracker-based sandboxing runtime that runs AI-agent tool calls inside isolated microVMs. Local-first and open source, with docs at emberd.hdprajwal.dev.
  • Agent-chaos: Chaos-engineering harness for LLM agents that injects deterministic faults at four well-defined seams. (work in progress)
  • QuackCode: Linux-first workspace for running several CLI coding agents (Claude Code, Codex, Gemini CLI) side by side per project, so you can hand off tasks between them without juggling terminal tabs or tmux.
  • Gitwise: CLI that turns staged diffs into commit messages and plain English into branch names.
  • Multimodal Malware Detection: Master's thesis comparing static-feature and visual-bytecode classifiers for malware detection.
  • Guardscribe: Real-time speech moderation pipeline that masks toxic spans under a 400ms end-to-end latency budget.

Side quests

things I built because I wanted to, not because anyone asked

  • Lucida: trace any image onto paper through your phone's camera.
  • Annoying UI: a gallery of intentionally hostile UI components.

Reach out on Twitter (@_hdprajwal) or email hdprajwal01 [at] gmail [dot] com.

Pinned Loading

  1. emberd emberd Public

    Firecracker-based sandboxing runtime that runs AI-agent tool calls inside isolated microVMs

    Python

  2. quackcode quackcode Public

    A personal AI coding agent, because I'm tired of waiting for official releases

    Rust

  3. multimodal-malware-detection multimodal-malware-detection Public

    Contains the implementation for my Master’s thesis on multimodal malware detection.

    Jupyter Notebook 1

  4. Guardscribe Guardscribe Public

    Real-time speech-to-text system with toxic content detection and filtering. Transcribes live audio using multiple ASR options while automatically detecting and masking harmful language.

    Python